Trials in adult critical care that show increased mortality of the new intervention: Inevitable or preventable mishaps?

Annals of Intensive Care

Table 2 Sample sizes, mortality rates, absolute risk increases and number needed to harm (NNH) in adult critical care RCTs showing increased mortality in the intervention groups

References	Sample size Total (intervention/control)	Mortality rates (n; %) (intervention/control)	Difference in number of deaths	P value	Absolute risk increases (%)	NNH
Hayes [10]	100 (50/50)	27/17 54/34 %	10	0.04	20	5
Fisher [15]	141 (108/33)	59/10 30/48/53/33 %^a (45/30)	49	0.02	15	6.6
Sloan [14]	98 (52/46)	24/8 46/17 %	16	0.03	29	3.5
Takala [13]^a	280 (119/123)	46/24 39/20 %	22	<0.001	19	5.3
Edwards [17]	10,008 (5007/5001)	1248/1075^a 25.7/22.3 %	173	0.0001	5	20
Perner [12]	798 (398/400)	203/172 51/43 %	31	0.03	8	12.5
Ferguson [16]	548 (275/273)	129/96 47/35 %	33	0.005	12	8.3
Finfer [8]	6030 (3016/3014)	829/750 27.5/24.9 %	79	0.02	2.6	38.5
Lopez [11]	797 (439/358)	259/175 59/49 %	84	<0.001	10	10
Gao Smith [9]	326 (162/164)	55/38 34/23 %	17	0.02	11	9
Totals or means	19,126 (9626/9462)	2879/2365 29.9/25 %	514		5	20

Four-week mortality rates of albumin versus control groups
In Fisher [15] RCT TNF-α receptor, the primary analysis was a trend test of placebo and increasing doses of TNF-α receptor. Mortality rates are shown for the dose groups and then for the combined intervention versus control groups (45/30)
^aSix-month mortality rates of methylprednisolone versus placebo